PSLE Science Reality Lab Vol No.171 | “These Data Are Provisional” — Does That Mean They Are Wrong?

PSLE-SCI-REALITY-0171

Wait, what? A number can be useful before it is final

You open a river-monitoring page and see a fresh graph. The newest readings arrived only minutes ago. Under the graph is a warning: These data are provisional and subject to revision.

One reader says, “Then the data must be wrong.” Another says, “It is on an official website, so it must already be final.” Both have pushed the label too far.

Provisional usually means the data are available early because timeliness is useful, but some review, correction, calibration check, field inspection or final approval may still happen. The readings are not automatically false. They are also not automatically final.

That middle position is exactly where good scientific reasoning lives: use the evidence, keep its status attached, and match the strength of the conclusion to the strength of the review completed so far.

Quick answer

If a scientific dataset is labelled provisional, a careful learner should read it as:

“This is the best currently available version for timely use, but it has not yet completed every review step and may be revised.”

Do not translate provisional into “wrong”, “fake”, “unusable”, “final”, “approved” or “guaranteed”. Instead ask what has already been checked, what is still pending, how large revisions could be, and what decision you are trying to make.

Owned learner job and boundary

Reality Lab Vol.171 owns one real-world evidence-transfer job: interpreting a provisional-data status label without treating it as a verdict on truth.

The broader PSLE Science skills remain with their existing owners. For method evaluation, use How to Evaluate PSLE Science Observations, Information and Methods. For updating an explanation when new evidence arrives, use the existing guide on revising explanations with new evidence. Reality Lab Vol.171 applies those skills to the status label learners see on real scientific dashboards.

The Night River case

Here is an original composite case. A river station sends a water-level reading every 15 minutes. At 9:00 pm the dashboard reports 2.38 m. At 9:15 pm it reports 2.44 m. At 9:30 pm it reports 2.51 m. A label says the real-time measurements are provisional.

A student writes in a group chat: “The river has definitely risen by 0.13 m in half an hour, because the official graph says so.” Another student replies: “You cannot use any of it because provisional means unreliable.”

A stronger scientific reading is more precise:

The current transmitted values show an apparent rise from 2.38 m to 2.51 m.
The values are provisional, so later review may revise them.
The short-term pattern can still be useful as timely evidence, especially for noticing change.
The reader should not pretend the exact values have completed final review.

Observed, claimed, inferred

Layer	Night River example
Observed / reported	The dashboard currently displays 2.38 m, 2.44 m and 2.51 m at the stated times.
Status information	The provider labels the data provisional.
Supported inference	The currently available record suggests the measured river stage increased over the interval.
Too strong	Every displayed value is final and can never change.
Also too strong	Because the data are provisional, none of the readings contain useful information.

Why scientific organisations publish provisional data at all

Some observations are valuable quickly. River level, rainfall, wave height, air measurements, earthquake data and weather observations may be useful soon after collection. Waiting months for every final review step would remove much of their timely value.

So a scientific organisation may publish an early version and clearly label its status. That is not a weakness in communication. It can be a sign of good provenance: the provider is telling you where the data are in their lifecycle.

The key is that speed and finality are different properties. A fast result can be useful without being final. A final result can be carefully reviewed but arrive long after the event.

What might change during later review?

Depending on the scientific system, later review may discover or account for things such as:

instrument malfunction;
sensor drift;
debris or changing physical conditions around a measuring station;
communication errors;
incorrect timestamps;
calibration information collected later;
field measurements that improve a conversion or rating relationship;
quality flags that were unavailable at the first transmission;
duplicate records or missing records;
processing rules that require a corrected value.

None of these possibilities tells you that a particular provisional value will change. They explain why the provider preserves the possibility of revision.

Representation check: the line on the graph can look more final than the data are

A smooth, solid line can create a strong visual feeling of certainty. The graph may use the same line style for provisional and final observations. Unless the status label is read, a learner can mistake visual neatness for review status.

Ask three representation questions:

Which points are provisional?
Does the graph visually distinguish provisional from approved data?
If not, where is the status explained—in a note, legend, metadata panel or disclaimer?

The scientific meaning may live outside the plotted line.

Comparison check: provisional versus approved is not the same as wrong versus right

Imagine that a provisional river value is 2.51 m and the later approved value becomes 2.49 m. The provisional value was revised by 0.02 m. It was not therefore “completely wrong”. It was an early measurement that became more carefully resolved after review.

Now imagine another point changes from 2.51 m to 1.90 m because a sensor malfunction was discovered. That revision is much larger. The label provisional does not tell you in advance how large a correction will be. Historical revision patterns, quality-control information and the measurement system matter.

The decision changes how much review you need

The same provisional dataset can be adequate for one purpose and inadequate for another.

Question	Possible use of provisional data
Is the river apparently rising right now?	Timely provisional observations may be highly useful, with the status kept visible.
What was the exact official annual maximum for a scientific report?	Wait for reviewed/approved data if the exact final value matters.
Should we compare two decades of long-term records?	Use a consistent reviewed dataset where possible and check version/status.
Is there an urgent signal worth checking?	Provisional data may help identify a signal, but confirmation and context still matter.

Worked case 1: the weather station correction

A school weather station automatically uploads temperature every five minutes. At 2:00 pm it displays 39.8°C, much higher than nearby readings. The value is provisional. The next day, the station operator finds that direct sunlight reached the sensor housing after a screen door was left open and flags the reading as invalid.

Before review: 39.8°C is a reported provisional observation and a reason to investigate.

After review: the evidence supports removing or qualifying that point because a method problem affected the measurement.

The important habit is not “trust provisional” or “distrust provisional”. It is keep the evidence status attached and update the conclusion when better evidence arrives.

Worked case 2: no revision at all

A wave buoy reports 1.8 m provisionally. Months later the reviewed archive still reports 1.8 m for that observation period. The fact that it was once provisional does not make the final value suspicious. It simply records that the observation passed through an earlier stage before review.

Worked case 3: a small change that matters to a threshold

A monitoring rule uses a threshold of 10.0 units. A provisional result is 10.1. After final quality review it becomes 9.9. The change is small, but the classification changes from above to below the threshold.

This teaches an important point: the importance of a revision depends not only on its size but also on the claim being made. A 0.2-unit change may be minor for one question and decisive for another.

Method check: what review is still pending?

When you see provisional, look for the provider’s explanation rather than inventing one. Useful questions include:

Has the instrument completed automatic quality screening?
Has a scientist reviewed the record?
Have field inspections been incorporated?
Has calibration been confirmed?
Is the conversion from a measured signal to the reported quantity final?
Will the provider replace provisional values with an approved archive?
How does the provider identify revisions?

Alternative explanations for a surprising provisional value

A surprising point could represent a real event, a sensor problem, a timing issue, an unusual local condition, a processing artefact or a communication error. The correct response is not to pick the explanation you like. It is to ask what additional evidence could discriminate among them.

Do nearby instruments show a similar event?
Does a later field inspection support the reading?
Are quality flags normal?
Does the value fit the instrument’s operating range?
Does the same signal appear in another independent measurement?
Does the provider later revise the point?

Evidence that strengthens confidence

Clear provenance and timestamps.
Automatic quality screening with transparent flags.
Agreement with independent measurements where appropriate.
Stable values after later review.
A documented review process.
Published revision history.
A later approved dataset that preserves the same scientific pattern.

Evidence that should make you narrow the claim

An instrument known to be malfunctioning.
A large unexplained jump exactly when equipment changed.
A provider warning that the current value is especially uncertain.
Missing calibration or site information.
Major later revisions to nearby observations.
A conclusion that depends on a tiny difference close to a threshold.

How far can a provisional conclusion travel?

A live value at one river station supports a claim first about that station, that time and the currently processed measurement. It does not automatically describe every point along the river, prove a long-term trend or establish the final annual record.

Scientific humility is not saying “we know nothing”. It is saying exactly what the current evidence supports and leaving room to update.

PSLE-style transfer case: the school pond logger

A school pond sensor uploads dissolved oxygen readings throughout the day. At noon the live dashboard shows 4.8 units and labels the data provisional. A classmate says, “We should throw away the reading because provisional means wrong.” Another says, “We can state 4.8 as the final official value because the computer displayed it.”

A strong scientific response is: the current dashboard reports 4.8, so it is usable as the present observation if its provisional status is kept. However, the value may still be revised after quality review, so it should not be described as final. If the exact value is important, the learner should check the later reviewed record.

Tempting but invalid reasoning

“Provisional means wrong.” Status is not a verdict on truth.
“Official website means final.” Official providers can deliberately publish clearly labelled preliminary data.
“If one point was revised, all provisional data are useless.” One revision does not define every observation.
“If most provisional values stay unchanged, no review is needed.” Review still matters, especially near thresholds or during instrument problems.
“A later revision proves dishonesty.” Revision can be the normal result of quality control and better evidence.

Model and measurement limits

Many modern scientific products are not simply “a sensor number”. A final record may combine raw measurements, calibration relationships, corrections, quality flags and expert review. A provisional value may therefore be an early stage in a processing chain rather than a finished scientific object.

This is another reason not to use one word—provisional—as a shortcut for the whole evidence system.

Delayed independent return

A real-time sensor value is provisional. State one thing you may still use it for and one claim you should avoid making.
A provisional reading changes only slightly after review. Was the first reading necessarily useless?
A provisional point is far above all nearby sensors. Name two different explanations and one piece of evidence that could help distinguish them.
A final annual statistic is needed for a report. Which version of the data should you prefer?

Explained answers

1. It may be useful for noticing a current pattern or change, but avoid calling it the final reviewed value.

2. No. It was an early estimate that happened to remain close to the reviewed value.

3. A real local event and a sensor problem are two possibilities. Nearby instruments, quality flags, field checks or later review could help discriminate.

4. Prefer the reviewed or approved dataset appropriate to the reporting purpose, while documenting the version used.

For parents and tutors: replace “trust/don’t trust” with “what stage?”

When a learner sees a scientific status label, avoid turning the lesson into a generic lecture about trustworthy websites. Stay inside the science. Ask:

What was measured?
How quickly was it published?
What checks have already happened?
What checks may still happen?
What conclusion is safe now?
What conclusion should wait for the reviewed record?

The habit you want is updateable confidence, not automatic belief or automatic rejection.

Why this belongs in PSLE Science

The current 2026 PSLE Science assessment objectives ask learners to interpret and analyse information, evaluate observations, information and methods, and communicate explanations and reasoning. The 2023 Primary Science syllabus also emphasises scientific inquiry, healthy scepticism, assumptions and uncertainty, more than one plausible explanation, evidence-based model building and understanding how Science is communicated in different forms and media.

A provisional-data label is therefore not an adult-only technical detail. It is a compact real-world invitation to ask: What do we know now, how was it obtained, what is still uncertain, and what could make us revise the conclusion?

Authoritative sources

Quiet return

Good science often arrives in stages. A provisional value can be useful evidence today and still be open to revision tomorrow. Keep the status attached, avoid pretending uncertainty is certainty, and avoid pretending uncertainty is ignorance. Use what the evidence can support now—and be ready to update when the reviewed record arrives.

Small Group Tutorials